How slow? The following shows that it takes an average of 0.7/1000th of a second to dedup a 4k line:

#! perl -slw
use 5.010;
use strict;
use Time::HiRes qw[ time ];;

sub uniq{ my %uniq; undef @uniq{ @_ }; keys %uniq }

my @lines;
for ( 1 .. 1e3 ) {
    my $line = int( rand 1e6 );
    $line .= chr(9) . int( rand 1e6 ) while length( $line ) < 4096;
    push @lines, $line;
}

my $start = time;
$_ = join chr(9), uniq( split chr(9), $_ ) for @lines;
my $stop = time;

printf "On random 4k lines, uniquing averaged %.6f seconds/line\n",
    ($stop - $start) / 1e3;

__END__
c:\test>junk
On random 4k lines, uniquing averaged 0.000695 seconds/line

c:\test>junk
On random 4k lines, uniquing averaged 0.000696 seconds/line

c:\test>junk
On random 4k lines, uniquing averaged 0.000712 seconds/line

c:\test>junk
On random 4k lines, uniquing averaged 0.000700 seconds/line
[download]

Which means processing your 3e6 lines should take around 35 minutes (plus a bit for the file processing).

So how fast do you want it to run? Or more likely, what are you doing in your full code that is slowing things down?